摘要
水质监测无人船对问题水域进行监测时,由于地形或天气原因使工作人员无法在视野中对水质监测无人船实时操作,需要USV的自主路径规划到指定水位进行检测。针对以上问题,本文提出利用无人船在未知水域中获得障碍物的分布信息,通过用Q学习算法对数据训练以规划路径,再利用BP神经网络的反馈进行权值的调整得到奖励值R,反馈给Q学习算法进行Q值迭代,其过程中选择不同的动作方向使Q值达到最优,从而使路径达到最优。最后通过实验仿真验证了该算法收敛速度更快,有效地提高路径规划效率,证明了该无人船路径规划算法的可行性。
When the water quality monitoring USV monitors the problem waters,due to terrain or weather reasons,the staff cannot monitor the water quality monitoring of the USV in real time.The USV's autonomous path planning is required to detect the water level.In view of the above problems,this article proposes to use the USV to obtain the distribution information of obstacles in the unknown waters.By using the Q learning algorithm to train the data for planning the path,the BP neural network feedback is used to adjust the weight to get the reward value R.The feedback is given to the Q learning algorithm for Q-value iteration,in which different action directions are selected to optimize the Q value,so that the path is optimal.Finally,the experimental results show that the algorithm converges faster and effectively improves the path planning efficiency.The demonstrations in this paper prove the feasibility of the USV path planning algorithm.
作者
吕扬民
陆康丽
王梓
LV Yangmin;LU Kangli;WANG Zi(College of Information Engineering,Zhejiang A&F University,Lin'an Zhejiang 311300,China)
出处
《智能计算机与应用》
2019年第1期14-18,23,共6页
Intelligent Computer and Applications
基金
浙江省重点研发计划项目(2015C0008)
关键词
水质监测无人船
路径规划
BP神经网络
强化学习
water quality monitoring USV
path planning
BP neural network
reinforcement learning